The document discusses the 'Eddie' middleware developed by the Eclipse Foundation's Oniro project, aimed at creating a unified view of virtual resources across devices and operating systems like Linux and Zephyr. It outlines the project's goals, including intelligent resource orchestration for seamless device functionality, illustrated by use cases such as transitioning video calls between devices. The presentation also covers the implementation architecture, communication methods, smart coordination strategies, and a roadmap for future developments focusing on ontologies and orchestration techniques.

1. SFSCon 2022, Bolzano
EDDIE - Enabling Distributed Intelligence
in Oniro
Stefan Schmidt <stefan.schmidt@huawei.com>
Principal Solution Architect, Huawei OSTC

2. Agenda
2
●
Introduction
●
Goals & Use Cases
●
Implementation: Linux & Zephyr
●
Roadmap
●
Summary

3. 3
Introduction

4. EDDIE
4
EDDIE – Enabling Distributed Intelligence
●
Eclipse Foundation - Oniro working group project
●
Cooperation with Politecnico di Milano
●
Middleware composed of different layers
●
Apache 2.0 license
●
C++ code base, constrained devices in mind
●
Implementation: Oniro Linux & Zephyr flavours
●
Work in progress

5. 5
Goals & Use Cases

6. UnifiedPoolofVirtualResources
6
“Objective to design a middleware capable of giving to the applications
and services an unified global view on cross-device resources. The
various devices are not seen any more as individual entities with specific
resources, but as a unified pool of virtual resources.”
●
Cross-device and also cross operating systems: starting with Linux and
Zephyr
●
For now resources as in peripherals and not general purpose resources like
memory and CPU

7. OrchestrationofAvailableResources
7
“Intelligent orchestration of the available resources will allow an efficient use,
as well as, some very unique use cases e.g. peripherals like camera, storage
and audio from any device in the network could operate like a local peripheral.
They would be instantly compatible with your local applications. ”
●
Even with many resources available its overwhelming to make good use of
them
●
At this stage the project goes from a simple rule-set based engine to complex
orchestration

8. ExampleUseCase:Videocall
8
●
Video call while on the commute back home
●
Call starts on your mobile
●
Enter your car: the unified pool of resources also has the car dashboard camera and audio
system avaiable and switches over
●
Walk from car to house: switching back to mobile as the car resources vanish
●
Enter kitchen: video moves to the kitchen TV or tablet, camera to the security camera and
audio to the multiroom audio in your house
Plethora of available, distributed, resources that can be used based on your personal
preferences and patterns.

9. 9
Implementation

10. Architecture
10
●
EDDIE is a middleware sitting on top of Oniro
●
Offering services and API to applications
●
Local agents part of every device
●
Virtual devices / super devices formed
out of local agents

11. Layers
11
●
Communication layer
●
Virtualization layer
●
Application management layer
●
Intelligent coordination layer
●
Forming the local agent

12. CommunicationLayer
12
●
Compared MQTT, Zenoh and CoAP for our use cases
●
Device and resource discovery
●
Constrained devices in mind (frame size, latency, memory, etc)
●
Settled on CoAP as it is fitting our purpose well
●
Libbcoap on Linux and a native CoAP implementation on Zephyr
●
IPv4 & IPv6
●
Running over low-power wireless connectivity
●
Resource Directory implementation on Linux

13. VirtualizationLayer
13
●
Resource Directory implementation turns out to work well for an initial resource
abstraction
●
Device announces itself via multicast to the network
●
If no other directory is found a Linux based local agent would elevate itself
●
Registers available resources at the found directory
●
API for upper layers to query by device type, attributes, URI’s, etc
●
Give uniform view on resources and access
●
Results are the base for decision making process

14. IntelligentCoordinationLayer
14
●
Current focus of the work
●
Research on options for ontologies/schemas to express attributes
●
Webthings, OCF, ROS, etc
●
Use ontologies to select the best execution environment
●
Different coordination approaches
– Centralized coordinator (current first approach)
– Any local agent could run as coordinator (with stand-by for failover)
– Complete peer-to-peer coordination, no central coordinator

15. UsageinOniroBlueprints
15
●
A blueprint to demonstrate how EDDIE can be used
●
Used between Linux and Zephyr
●
Used over OpenThread with IPv6 (low-power and loosy network)
●
M2M between touch controller and doorlock blueprints
●
Based on availability of resources the touch controller can adjust UX dynamically

16. 16
Roadmap

17. FutureRoadmap
17
2023
●
Focus on ontologies/schemas
●
Focus on intelligence orchestration
●
Explore different coordination approaches

18. ONIROPROJECT
Thank you!
Join us @
Eclipse Foundation booth
& oniroproject.org
31